Benefits of brain dual-energy CT imaging in detecting intracranial hemorrhage in noncontrast brain CT scans

doi:10.21203/rs.3.rs-2570940/v1

Background — Noncontrast brain computed tomography (CT) scans are the gold standard for diagnosing intracranial hemorrhage (ICH). However, the scans have some evaluation limitations, especially in cases of interfering bony artifacts, and overlapping hypoattenuating calcifications and hemorrhages. Material decomposition dual-energy CT (DECT) can separate materials based on atomic numbers. We compare the diagnostic accuracy, sensitivity, and specificity of material decomposition noncontrast DECT of the brain with stimulated conventional 120 kVp-single energy CT (sSECT) imaging for ICH diagnosis.

Methods — This retrospective study was conducted at a single center. Three neuroradiologists separately and blindly reviewed the postprocessing images of 111 patients with 215 lesions. DECT acquisitions generated the images. They were in the form of:

♦ sSECT images,

♦ blood-subtracted calcium, nonoverlay images (“blood [calcium] nonoverlay” images),

♦ blood-subtracted calcium, overlay images with rainbow and grayscale color-coding (“blood [calcium] overlay” images), and

♦ combined sSECT and blood-subtracted calcium, overlay images (“combined sSECT + blood (calcium) overlay” images).

Results — Compared with the sSECT images, the blood (calcium) overlay images had better accuracy (99.6%), sensitivity (99.5%), and specificity (100%) for ICH diagnosis. The blood (calcium) overlay images also provided better observer confidence (a 94.06% certain diagnosis) than the sSECT images (P < 0.05). The blood (calcium) overlay images and combined images had very good interrater reliability, whereas the reliability of the sSECT images was poor.

Conclusions — Material decomposition DECT—as either blood (calcium) overlay images or combined sSECT + blood (calcium) overlay images—has a high ICH diagnostic ability in noncontrast brain CT imaging.

Dual energy CT

noncontrast CT brain

intracranial hemorrhage

ICH

Small intracranial hemorrhage nearby high attenuation cranial structures such as bony calvarium, calcification can miss on conventional noncontrast CT brain
Dual energy CT (DECT) have capilities in differentiation hemorrhage, bone or calcification from each other
Blood(calcium) overlay images shows high accuracy, sensitivity, specificity in diagnosis ICH as well as highest confident interobsever reliability in diagnosis ICH
Radiation dose by DECT is similar to conventional CT brain

Head injury is a severe problem in all countries due to its high morbidity and mortality. The incidence of head injury varies depending on the terminology used and the data collection method. In 2013, traumatic brain injuries were diagnosed in nearly 2.8 million of the 26 million injury-related emergency department visits, hospitalizations, and deaths that occurred in the United States. [1] In Thailand, 42.7% of severe injury cases related to motorcycle accidents had head injuries, and the corresponding figure in cases of death was 70.8%. [2]

Computed tomography (CT) interpretation of intracranial hemorrhage (ICH) is based on attenuation in biological tissues. The attenuation of the tissue is associated with its atomic number and physical density. Magnetic resonance imaging (MRI) can provide a more accurate than CT in stage of the hematoma by identifying the sequential patterns of transformation of hemoglobin within the hematoma. [3] However, MRI has limited availability and requires a long scan time. Thus, noncontrast CT remains the diagnostic test of choice in acute intracerebral hemorrhage because of its sensitivity to acute hemorrhage, short scan time, low cost, widespread availability, and the technical ease of obtaining a scan even with acutely ill patients. [4] Nevertheless, CT scans still have limitations in evaluating small lesions and lesions close to the skull base and posterior fossa. Such lesions may interfere with bony artifacts and can easily be misdiagnosed. [5]

Hyperattenuated materials in noncontrast CT scans can arise from 2 common causes: hemorrhage and calcium. The other less common causes are high proteinaceous substances and hypercellular tissue. In noncontrast CT images, acute hemorrhage has an attenuation range of 60–100 Hounsfield units (HU) [6], whereas calcified lesions generally have an attenuation > 100 HU. However, calcification can also have varying attenuations; these can be less than 100 HU, depending on the amount of calcium deposition. Therefore, there is an overlap between the attenuation values of low attenuation intracranial calcification (< 100 HU) and hemorrhage, and they also show the blooming effect of the SWI sequence. These shortcomings may lead to misinterpretations [7] and consequential inappropriate management.

A dual-energy CT (DECT) scan is a new CT scanning technology that generates material-specific images based on a particular material’s atomic number and unique mass attenuation coefficient at different X-ray energies. Material-specific images provide qualitative and quantitative information on tissue composition. [8] Moreover, DECT imaging is not associated with increased radiation dose levels. [9] Therefore, DECT scans are ubiquitously used for virtual monochromatic images, artifact suppression images, and material decomposition images. According to brain tissue, blood, and calcium/bone have different atomic numbers, modified from Yang, 1987. [10]. The effective atomic number of blood (atomic number = 7.59) is different from that of calcium (atomic number = 20.0). Consequently, DECT may provide improved diagnostic accuracy, sensitivity, and specificity in detecting ICHs in noncontrast DECT of the brain.

The primary research purpose was to determine the diagnostic accuracy, sensitivity, and specificity of noncontrast DECT of the brain in diagnosing any ICH compared with stimulated conventional 120 kVp-single energy CT (sSECT) imaging. The secondary aims to collect the radiation dose and estimate the total scan time of noncontrast DECT imaging.

This retrospective study drew upon the picture archiving and communication system (PACS). After approval by the institutional ethics committee, a review was made of all consecutive, noncontrast, DECT examinations of the brain performed at Siriraj Hospital between May 1, 2019, and January 31, 2020.

Study Participants

Inclusion criteria

The inclusion criteria were as follows:

The patient underwent noncontrast DECT of the brain at Siriraj Hospital between May 1, 2019, and January 31, 2020.
The patient was > 18 years old.
The patient had a confirming diagnosis noted by clinical information in at least one of the following information such as the outpatient department (OPD) card records, operative findings, or records related to a follow-up CT or MRI of the brain performed for any reason within 1 month after the first noncontrast DECT imaging.
In the “negative group” the patient was approved for discharge by a clinician with “clinically negative for ICH” noted in the OPD card records. If there was a follow-up appointment, the patient did not present abnormal neurological symptoms that may have led to suspicion of ICH (such as nausea, vomiting, headache, or an abnormal neurological examination).

Exclusion criteria

The exclusion criteria were as follows:

The patient did not undergo a noncontrast DECT scan of the brain at Siriraj Hospital.
The patient did not have a clinically confirmed diagnosis, either through CT or MRI of the brain performed within 1 month or via intraoperative findings.
The postprocessing image from the DECT scan was of poor quality, or the related data file was incomplete.
The patient underwent an intracranial operation less than 1 year before the DECT was performed.

Image Acquisition And Postprocessing

The noncontrast DECT of the brain was performed at our institution with a 256-slice, single-source, tube voltage switching, DECT scanner (Revolution CT scanner; GE Healthcare, Chicago, IL, USA) with a helical scan. The tube voltage was switched between 80–140 kV, while the tube current modulation was fixed at approximately 315 mA. Primary imaging reconstruction was performed at 65 keV.

The retrospective postprocessing reconstruction images were obtained with material decomposition algorithm at 65–75 keV to maximize imaging quality, according to Pomerantz 2012. [11] This involved the axial view, a 2 mm-slice thickness without interval gap, a 50–104 window width, and a 1000–1040 window level. There were 4 data sets of images: 1) sSECT images; (2) blood-subtracted calcium, nonoverlay images (“blood [calcium] nonoverlay” images); (3) blood-subtracted calcium, overlay images with rainbow and grayscale color-coding (“blood [calcium] overlay” images); and (4) combined sSECT images and blood-subtracted calcium, overlay images (“combined sSECT + blood [calcium] overlay” images).

Data analysis

Demographic data

The demographic data of the enrolled patients were reviewed by 1 radiologist independently from the neuroradiologist readers. The following demographic data were obtained from electronic OPD card records:

Patient age
Patient sex: male/female
Cause of injury: falling, traffic accident, other
Glasgow Coma Scale (GCS) score
Final clinical diagnosis of ICH: type, stage, location
Clinical management: discharged, referred out, admitted for conservative treatment or for observation of their neurological signs and symptoms, surgery, death, admitted due to a cause other than ICH, discharged against advice
Confirmation of diagnosis: follow-up CT imaging of the brain within 1 month (follow-up CT scanning within 48 hours or > 48 hours), follow-up MRI of the brain within 1 month, follow-up clinical context, surgery with a positive operative finding of ICH, death with/without an autopsy
Radiation dose: CT dose index volume (CTDI_vol; mGy), total dose-length product (mGy-cm)
Scan time (msec)

Image analysis

Postprocessing of the DECT images of the enrolled patients was performed by one in the research team. It was conducted independently from the neuroradiologist readers. After that, the postprocessing images were obtained as 4 sets of images (Fig. 1):

sSECT images
blood-subtracted calcium, nonoverlay images (blood [calcium] nonoverlay images)
blood-subtracted calcium, overlay images with rainbow and grayscale color-coding (blood [calcium] overlay images)
sSECT co-interpreted with blood (calcium), overlay images (combined sSECT + blood [calcium] overlay images)

Lesions that still appeared hyperattenuating in blood (calcium) nonoverlay images were classified as hemorrhagic lesions. However, lesions that showed isoattenuation or hypoattenuation in blood (calcium) nonoverlay images were classified as nonhemorrhagic tissues. In blood (calcium) overlay images, lesions highlighted in yellow-green with rainbow color-coding or appearing hyperattenuated with grayscale color-coding were classified as hemorrhagic lesions. Lesions with any other color with rainbow color-coding or lesions with isoattenuation or hypoattenuation (grey-black with the grayscale color-coding) were classified as nonhemorrhagic (Fig. 1; Table 1).

Table 1 Criteria used for classification of lesion

Type of lesion	Appearance on each image set
	sSECT	Blood(calcium)	Blood(calcium) overlay
	sSECT	Blood(calcium)	Rainbow color code	Greyscale color code
Hemorrhagic lesion	hyperattenuating	hyperattenuating	yellow-green color	hyperattenuating
Non-hemorrhagic lesion	Iso-/hypo-attenuating	Iso-/hypo-attenuating	Any color (blue/orange/red color)	Iso- or hypoattenuating (grey-black color)

Table 2 Demographic data shows patient and disease

characteristics.

	Positive group (n = 58)	Negative group (n = 53)	p-value
Age [mean±SD]	49.40 ± 22.50	64.74 ± 20.22	< 0.001
Gender (number, %) Male	39 (67.2%)	27 (50.9%)	0.081
Female	19 (32.8%)	26 (49.1%)
Cause of injury (number, %) Falling	20 (34.5%)	38 (71.7%)	< 0.001
Traffic	31 (53.4%)	7 (13.2%)
Other	7 (12.1%)	8 (15.1%)
GCS score [Median, (min, max)]	15 (2T, 15)	15 (6, 15)	< 0.001
Clinical Management (number, %) Discharged	2 (3.4%)	29 (54.7%)	< 0.001
Referred out	8 (13.8%)	1 (1.9%)
Admitted for conservative management	33 (56.9%)	2 (3.8%)
Admitted due to other cause	5 (8.6%)	21 (39.6%)
Surgery	6 (10.3%)	0 (0%)
Against advised	1 (1.7%)	0 (0%)
Death	3 (5.2%)	0 (0%)
Confirmation of Diagnosis (number, %) F/U¹ brain CT scan ≤ 1 month	24 (41.4%)	7 (13.2%)	< 0.001
F/U brain CT scan ≤ 48 hours	17 (29.3%)	4 (7.5%)
F/U brain CT scan > 48 hours – 1 month	7 (12.1%)	3 (5.7%
F/U brain MRI ≤ 1 month	2 (3.4%)	1 (1.9%)
F/U clinical context	24 (41.4%)	45 (84.9%)
Surgery with positive ICH²	6 (10.3%)	0 (0%)
Death with/without autopsy	2 (3.4%)	0 (0%)

GCS = Glasgow coma scale/score, F/U = follow-up, ICH = intracranial hemorrhage

All images were interpreted independently by three neuroradiologists (with more than 8 years of experience) who were blinded to the patients’ clinical contexts, any prior or follow-up imaging, and other observers’ reports. Multifocal ICHs of the same type in the same area were recorded as 1 lesion to avoid overcounting nonindependent lesions (such as multifocal hemorrhages). Multifocal ICHs of different types or from different areas in the same patient were recorded as separate lesions.

Each set of images was reviewed separately at least 2 weeks after the initial assessment, with blinding to the prior report. Each observer gave each lesion a confidential reporting score. The scores were 2 for a “certain diagnosis,” 1 for a “probable diagnosis,” and 0 for a “doubtful diagnosis.”

The data were recorded, and the lesion subgroups were categorized by a radiology resident as follows:

1. ICH

1.1 Positive subgroup. Patients in this subgroup were classified in this manner:

Type of ICH (5): epidural hemorrhage (EDH), intraparenchymal hemorrhage (IPH), intraventricular hemorrhage (IVH), subarachnoid hemorrhage (SAH), and subdural hemorrhage (SDH)
Stage of ICH (3): acute, subacute, and chronic
Size. The sizes of measurable lesions (EDH, IPH, and SDH types) were recorded at the maximal thickness (cm) for EDH and SDH and the maximal dimension (cm) for IPH. The lesions were classed as small (< 1 cm) or large (> 1 cm).
Location
By area: the falx-tentorium-parasagittal area, cerebral hemisphere, base/inferior brain adjacent to the skull base, posterior fossa, intraventricular space, and other subarachnoid space not along sulci and gyri of the cerebral hemisphere.
By lobe: frontal, parietal, temporal, occipital, brainstem, and cerebellum
By side: left, right, and bilateral

By IPH location: deep and lobar

1.2 Negative subgroup

The final imaging diagnosis was reviewed from the concordant diagnosis between 2 readers. If there was an asynchronous diagnosis, the third reader was reviewed, and a consensus was reached to obtain the final diagnosis. The reference standard of negative or positive hemorrhage was established by consensus diagnosis. The diagnosis was correlated with commonly used and relevant prior or follow-up CT or MRI imaging and the available clinical information. Intraoperative findings were also used to determine the reference standard when available.

Furthermore, in the positive-ICH group, when a subsequent CT scan of the brain within 48 hours after the DECT imaging was recorded, the initial and follow-up lesion size, initial and follow-up lesion HU in noncontrast imaging.

Statistical analysis

Categorical independent variables (such as sex, cause of injury, clinical management, and confirmation of diagnosis) are presented as numbers or percentages. Continuous data (for instance, age, radiation dose, and scan time) are reported as the mean ± SD. The GCS scores are reported as the median (minimum, maximum) as an appropriated distribution.

The associations for univariate analysis and the categorical independent variables of the positive and negative subgroups (age, sex, cause of injury, GCS score, clinical management, and confirmation of diagnosis) were compared.

Diagnoses

The diagnostic performance of the sSECT images, blood (calcium) nonoverlay images, blood (calcium) overlay images, and combined sSECT + blood (calcium) overlay images was reported by accuracy, sensitivity, and specificity with 95% CIs. Subgroup analyses (such as type, stage, size, and location) were performed by the Chi² test. The significance of continuous variables as two-independent samples was assessed with independent t-tests or Mann–Whitney U tests.

Interrater reliability was carried out by Cohen’s kappa and 95% CIs. A probability (P) value less than 0.05 was considered statistically significant.

Our retrospective review of the data stored in the hospital’s PACS found that 841 noncontrast DECT studies of the brain were performed at our institution between May 1, 2019, and January 31, 2020. We excluded 730 studies as they failed to meet the inclusion criteria. Of those, 689 had incomplete PACS data files; 38 did not have neurological signs or symptoms noted in their electronic OPD records; 1 involved a recent intracranial operation; and 2 had more than 7 intracranial hemorrhagic foci. Thus, 111 studies with 251 lesions were reviewed (58 positive studies with 198 lesions and 53 negative studies with 53 lesions.

Demographic Data

The study enrolled 111 patients (male 59.5%; female 40.5%). The mean age of positive-ICH patients (49.40 ± 22.50 years) was significantly lower than that of negative-ICH patients (64.74 ± 20.22 years). The minimum GCS score was 2T for positive-ICH patients and 6 for negative-ICH patients (Table 2). The major cause of injuries was falling (52.3%). It was followed by traffic accidents (34.2%) and other causes (13.5%), such as body assault and acute alteration of consciousness with suspected ICH

Most positive-ICH patients were admitted for observation of their neurological signs and symptoms (56.9%), follow-up brain CT imaging (41.4%), or follow-up for other clinical reasons (41.4%). Seventeen positive-ICH patients (29.3%) were followed up with a CT scan of the brain within 48 hours. Conversely, most negative-ICH patients were discharged from the outpatient department (54.7%) and had a follow-up session (84.9%; Table 2).

Diagnostic Performance For Ich

This study revealed that sSECT images had a 79% diagnostic accuracy, 76.8% sensitivity, and 90.6% specificity (Table 3). SSECT images are equivalent to the conventional single-energy CT scans commonly used as the gold standard for diagnosing ICH.

Table 3

Diagnostic performance of the images set in diagnosis of ICHs, evaluated per lesion (n = 251 lesions)
Test set	%Accuracy (95%CIs)	%Sensitivity (95%CIs)	%Specificity (95%CIs)
sSECT	79.7 (74.2, 84.5)	76.8 (71.1, 81.9)	90.6 (86.3, 93.9)
Blood(Calcium)	61.0 (54.7, 67.1)	56.6 (50.2, 62.8)	77.4 (71.7, 82.4)
Blood(Calcium) overlay	99.6 (97.8, 100.0)	99.5 (97.6, 100.0)	100 (98.5, 100.0)
sSECT + Blood(Calcium)	92.8 (88.9, 95.7)	90.9 (86.6, 94.2)	100 (98.5, 100.0)

The blood (calcium) overlay images had the highest diagnostic accuracy (99.6%), highest sensitivity (99.5%), and highest specificity (100.0%). All values were significantly higher than those of sSECT images (P < 0.0001).

In contrast, the blood (calcium) nonoverlay images had the lowest diagnostic accuracy (61.0%), lowest sensitivity (56.6%), and lowest specificity (77.4%). These values were significantly lower than those of sSECT images (P < 0.0001).

However, co-interpretation of the combined sSECT + blood (calcium) overlay images increased diagnostic accuracy to 92.8%, sensitivity to 90.9%, and specificity to 100.0% (P < 0.0001).

Subgroup analysis of the 198 positive lesions for the 5 ICH types (EDH, IPH, IVH, SAH, and SDH) revealed that blood (calcium) overlay images had the highest diagnostic accuracy for all types (Figs. 2 and 3; Table 4).

Table 4

Diagnostic performance of the images set in diagnosis of ICHs, evaluated per lesion: Subgroup analysis of the positive lesions (n = 198 lesions) via types of ICHs
Test set	Type of ICH [n = 198]					p-value
Test set	IPH [n = 29]	SDH [n = 93]	EDH [n = 17]	SAH [n = 50]	IVH [n = 9]	p-value
sSECT	25 (86.2%)	67 (72%)	15 (88.2%)	37 (74.0%)	8 (88.9%)	.306
Blood(Calcium)	18 (62.1%)	54 (58.1%)	15 (88.2%)	18 (36.0%)	7 (77.8%)	.001
Blood(Calcium) overlay	29 (100.0%)	93 (100.0%)	17 (100.0%)	49 (98.0%)	9 (100.0%)	.562
sSECT + Blood(Calcium)	26 (89.7%)	88 (94.6%)	17 (100.0%)	41 (82.0%)	8 (88.9%)	.086
ICH = intracranial hemorrhage, IPH = intraparenchymal hemorrhage, SDH = subdural hemorrhage, EDH = epidural hemorrhage, SAH = subarachnoid hemorrhage, IVH = intraventricular hemorrhage

The sizes of the 139 measurable lesions (EDH, IPH, and SDH) found by this study were measured in the axial view and at the maximal dimension. In the case of EDH, the mean size was 2.55 (SD 4.15) cm. The thinnest EDH was 0.46 cm and was detected in all image sets (Table 5). There were no statistically significant differences between the image sets (P = 1.0000).

Table 5

Diagnostic performance of the images set in diagnosis of ICHs, evaluated per lesion: Subgroup analysis of the positive lesions (n = 198 lesions) via stages of ICHs
Test set	Stage of ICH [n = 198]			p-value
Test set	Acute [n = 191]	Subacute [n = 5]	Chronic [n = 2]	p-value
sSECT	145 (75.9%)	5 (100.0%)	2 (100.0%)	.334
Blood(Calcium)	108 (56.5%)	2 (40.0%)	2 (100.0%)	.351
Blood(Calcium) overlay	190 (99.5%)	5 (100.0%)	2 (100.0%)	.982
sSECT + Blood(Calcium)	174 (91.1%)	4 (80.0%)	2 (100.0%)	.629

The mean IPH size was 2.09 (SD 1.66) cm. The thinnest IPHs detected in the sSECT, blood (calcium) nonoverlay, blood (calcium) overlay, and combined sSECT + blood (calcium) overlay images were 0.39, 0.39, 0.27, and 0.27 cm thick, respectively (Table 5). No statistically significant differences were found between each set of images (P = 0.7841).

As for SDH, the mean size was 0.44 (SD 0.35) cm. The thinnest SDH lesion was 0.12-cm thick and located along the falx cerebri. It could be detected only in blood (calcium) overlay images and the combined sSECT + blood (calcium) overlay images. The thinnest SDH lesion visualized in sSECT images was 0.14-cm-thick; it was along falx cerebri. Blood (calcium) nonoverlay imaging detected an SDH lesion as small as 0.17-cm thick in the left frontal area (Table 6). However, there were no statistically significant differences between each set of images (P = 0.3313–0.6972).

Table 6 Maximum and minimum size of the measurable ICHs, evaluated per lesion: Subgroup analysis of the positive lesions (n=139 lesions) via type of the measurable ICHs

Diagnoses Confidence Levels

The observers had the highest confidence in their ICH diagnosis when using blood (calcium) overlay images (94.1% of diagnoses were rated “certain diagnosis”). This confidence level was statistically significantly higher than the corresponding proportions for combined sSECT and blood (calcium) overlay images (P < 0.05; Fig. 4). However, there was no statistically significant difference from combined sSECT + blood (calcium) overlay images).

Interrater Reliability

The interrater reliability was assessed for each image type. The ICH diagnosis was lowest when using sSECT images (Cohen’s kappa = 0.38), which indicated poor agreement. Using blood (calcium) nonoverlay and blood (calcium) overlay images improved interrater reliability to good agreement (Cohen’s kappa = 0.67 and 0.63, respectively).

Quality Of Dect Imaging Assessments

The mean CTDI_vol per study was 53.64 (SD ± 4.59) mGy, and the mean total dose-length product was 1103.30 (SD ± 188.05) mGy-cm, which was equivalent to an effective dose of 2.32 (SD ± 0.39) mSv. The mean DECT scan duration was 53.92 (SD ± 19.16) seconds.

Head injury is a severe problem in most countries worldwide and causes high morbidity and mortality. Noncontrast CT imaging is the gold standard for ICH diagnosis. However, it has limitations with the evaluation of small lesions and with those located close to the skull base and the posterior fossa, where bony artifacts can cause interference. [5] The common hyperattenuating materials in noncontrast CT scans are hemorrhage and calcium. However, there is overlap attenuation between hemorrhage and low attenuation calcification. Nevertheless, hemorrhage (blood) and calcium have individual atomic numbers that can be used to differentiate them in material decomposition DECT imaging. This distinction is advantageous because hemorrhage and calcification require different management approaches.

This study enrolled 111 patients without different sex proportions between the positive- and negative-ICH groups. The mean age of the positive-ICH patients was significantly lower than that of the negative-ICH patients. The common cause of injury was a traffic accident. Such injuries are high energy and commonly occur in young, active-lifestyle patients. Additionally, the minimum GCS score of the positive-ICH group was lower than that of the negative-ICH group. This difference implies greater injury severity among the patients in the positive-ICH group. However, most of the positive-ICH patients were admitted for observation of their neurological signs and symptoms with a follow-up brain CT scan or some other clinical reason. Conversely, most negative-ICH patients had a higher minimum GCS score, and they may have been discharged without a follow-up appointment, given that they had less severe symptoms.

The results of this study support the advantages of DECT scans in differentiating calcium from hemorrhage in phantom models [9, 12] and human models [13, 14], given its improved diagnosis of ICH in noncontrast CT images of the brain. However, only blood (calcium) overlay images and the combined sSECT + blood (calcium) overlay images increased diagnostic accuracy, sensitivity, and specificity compared with sSECT images alone, which did not depend on ICH type, stage, size, or location. The blood (calcium) overlay images also increased the observers’ confidence levels and improved interrater reliability. Thus, material decomposition DECT can determine whether hyperattenuating lesions contain blood and minimize reader-related subjectivity (signified by the good interrater reliability). Moreover, the distinctive color of blood, especially in blood (calcium) overlay images, is easily detected by both neuroradiologists and non-radiologists. This attribute facilitates quick detection and decision-making about further management. The advantages of material decomposition DECT in discriminating between blood and calcium, correlating with clinical context, facilitate a decrease in unnecessary admissions and a reduction in the costs of hospitalizing negative-ICH patients.

On the other hand, this study found that the blood (calcium) nonoverlay images alone had inferior diagnostic accuracy, sensitivity, and specificity to sSECT images. This finding could result from the poor image quality of blood (calcium) nonoverlay images and the difficulty in discriminating the attenuation of the images. Nevertheless, the blood (calcium) nonoverlay images improved interrater reliability. This increase resulted from their diagnostic ability being higher for large ICHs than small ICHs, which meant that every reader could easily detect large ICHs in the images. Thus, we do not recommend using blood (calcium) nonoverlay images alone: they should be co-interpreted with sSECT images.

The sSECT images had the poorest diagnostic ability and the lowest interrater reliability, which would better detect large ICHs than small ones, and it was prone to subjective interpretation. Despite that, the sSECT images still had the highest quality image due to their good grey-white differentiation.

The mean CTDI_vol and effective radiation doses of DECT in this study were nearly equal to the radiation dose of SECT from a prior study. [15, 16, 17] The mean duration of the DECT scan per case was under a minute, which could be beneficial in an emergency setting; however, the postprocessing of images was slightly longer for DECT than for SECT.

Some limitations exist in this study. First is retrospective design, many samples were unavailable for analysis due to incomplete data in postprocessing image files. Second, MRI studies of the brain were not performed to confirm the diagnoses in all cases due to resource limitations. Given that, the reference standards of this study were based on consensus diagnoses correlated with commonly used relevant prior or follow-up CT or MRI imaging and clinical information. Third, the material decomposition DECT needs source data to process images. The data occupies a large data volume in a PACS, and the system needs to be stable for timely processing. Last, our study’s postprocessing images of overlay DECT could only be exported as screen-capture, which meant that multiplanar reconstruction could not be performed. Consequently, the observers had to interpret images only in the axial plane, which probably resulted in the misdiagnosis of some lesions.

Material decomposition DECT scans, especially blood (calcium) overlay images, increased the diagnostic accuracy, sensitivity, and specificity for ICHs in noncontrast brain CT scans and increased the observer confidence level and interrater reliability. Material decomposition DECT is helpful, especially with uncertain hyperattenuating foci that are unclear in conventional CT images alone. The distinctive color of blood in blood (calcium) overlay images can be easily detected even by non-radiologists. This attribute might facilitate quick decision-making about further management in emergency settings. Furthermore, correlating the clinical context with DECT’s ability to discriminate between blood and calcium could decrease unnecessary admissions and reduce the costs of hospitalizing negative-ICH patients. We recommend using noncontrast brain DECT scans, especially in emergency settings.

Ethical Approval : The Siriraj Institutional Review Board of our institution approved the protocol of the present study. The SIRB protocol number of the study is 194/2563 (IRB1). certify that the study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Consent to participate : Not applicable

Consent for publication : Not applicable

Availability of data and materials : The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests : The authors declare that they have no competing interests.

Funding : Not applicable

Authors' contributions: Pipat Chiewvit: PC

Kanyaphak Ratanakarn : KR

Siriwan Piyapittayanan : SP

Chanon Ngamsombat : CN

Tipa Chakorn: TC

Conceived the original idea and developed the theoretical formalism : PC , TC

Collect and analysis data : PC, KR

Aided in interpreting the results : SP, CN

Discussed the results, commented on the manuscript, and reviewed the manuscript: All wrote the main manuscript text: PC, KR approved the manuscript before submission, including the names and order of authors, and that all authors receive the submission and all substantive correspondence with editors, as well as the full reviews, verifying that all data, figures, materials (including reagents), and code, even those developed or provided by other authors, comply with the transparency and reproducibility standards of both the field and journal : PC

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST

The authors declare that they have no potential conflicts of interest related to the research and did not receive any financial support for the research, authorship, or publication of this article.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the contributions of Assist. Prof. Chulaluk Komoltri and Ms. Nerisa Thornsri, who performed the statistical analyses for this research. The authors are also indebted to Mr. David Park for English-language editing.

Peterson A. Surveillance Report of Traumatic Brain Injury-related Emergency Department Visits, Hospitalizations, and Deaths-United States, 2014. Centers Dis Control Prev US Dep Heal Hum Serv [Internet]. 2019;24. Available from: www.cdc.gov/TraumaticBrainInjury
Surveillance I. Thailand’s Report Situation of Severe Injuries Year 2005-2010. 2010;97.
Parizel PM, Makkat S, Van Miert E, Van Goethem JW, Van den Hauwe L, De Schepper AM. Intracranial hemorrhage: Principles of CT and MRI interpretation. Eur Radiol. 2001;11(9):1770–83.
Aygun N, Masaryk TJ. Diagnostic imaging for intracerebral hemorrhage. Neurosurg Clin N Am. 2002;13(3):313–34.
Bahner ML, Reith W, Zuna I, Engenhart-Cabillic R, Van Kaick G. Spiral CT vs incremental CT: Is spiral CT superior in imaging of the brain? Eur Radiol. 1998;8(3):416–20.
Günther M. Perfusion imaging. J Magn Reson Imaging. 2014;40(2):269–79.
Nute JL, Roux L Le, Chandler AG, Baladandayuthapani V, Schellingerhout D, Cody DD. Differentiation of low-attenuation intracranial hemorrhage and calcification using dual-energy computed tomography in a phantom system. Invest Radiol. 2015;50(1):9–16
Patino M, Prochowski A, Agrawal MD, Simeone FJ, Gupta R, Hahn PF, et al. Material separation using dual-energy CT: Current and emerging applications. Radiographics. 2016;36(4):1087–105.
Henzler T, Fink C, Schoenberg SO, Schoepf UJ. Dual-energy CT: radiation dose aspects. AJR Am J Roentgenol. 2012;199(5 Suppl):16–25.
Yang NC, Leichner PK, Hawkins WG. Effective atomic numbers for low-energy total photon interactions in human tissues. Med Phys [Internet]. 1987 Sep;14(5):759–66. Available from: http://dx.doi.org/10.1118/1.596000
Pomerantz SR, Kamalian S, Zhang D, Gupta R, Rapalino O, Sahani D V., et al. Virtual monochromatic reconstruction of dual-energy unenhanced head CT at 65-75 keV maximizes image quality compared with conventional polychromatic CT. Radiology. 2013;266(1):318–25.
Nute JL, Jacobsen MC, Chandler A, Cody DD, Schellingerhout D. Dual-Energy Computed Tomography for the Characterization of Intracranial Hemorrhage and Calcification: A Systematic Approach in a Phantom System. Invest Radiol. 2017;52(1):30–41.
Gupta R, Phan CM, Leidecker C, Brady TJ, Hirsch JA, Nogueira RG, et al. Evaluation of dual-energy CT for differentiating intracerebral hemorrhage from iodinated contrast material staining. Radiology. 2010;257(1):205–11.
Wiggins WF, Potter CA, Sodickson AD. Dual-energy CT to differentiate small foci of intracranial hemorrhage from calcium. Radiology. 2020;294(1):129–38.
Christner JA, Kofler JM, McCollough CH. Estimating effective dose for ct using dose-length product compared with using organ doses: Consequences of adopting international commission on radiological protection publication 103 or dual-energy scanning. Am J Roentgenol. 2010;194(4):881–9.
Almeida IP, Schyns LEJR, Öllers MC, van Elmpt W, Parodi K, Landry G, et al. Dual-energy CT quantitative imaging: a comparison study between twin-beam and dual-source CT scanners. Med Phys. 2017;44(1):171–9.
Coakley F V., Gould R, Yeh BM, Arenson RL. CT radiation dose: What can you do right now in your practice? Am J Roentgenol. 2011;196(3):619–25.

No competing interests reported.

Benefits of brain dual-energy CT imaging in detecting intracranial hemorrhage in noncontrast brain CT scans

Status:

Version 1

Abstract

Figures

Key Points

Background

Methods